This paper discusses certain problems that can arise with respect to the selection of content in the preparation of a test to measure language proficiency for some specific purpose. Forty employees in a defense plant in Montreal, Canada were interviewed in order to specify language needs, to arrive at a precise interpretation of the required reading standard, and to produce appropriate measures of reading proficiency. The main problem involved coming up with a common test that everyone could take and that would contain job-related content. An analysis of the general textual characteristics of work orders and manuals led to the identification of three linguistic skills as the overall testing objectives. Generally these skills involved understanding relationships between parts of sentences and paragraphs, and the ability to identify specific linguistic elements whose characteristics would correspond to the characteristics asked for in a given question. Analysis of the lexical material excluded items that were truly technical. Two conclusions are that the special purposes approach to assessment appears most appropriate when the examinees are as homogeneous as possible, and very accurate definitions of "technical" and "special" are required. An appendix contains examples of item types used. (AMH)